Design and Implementation of REST API for Predicting the Recitation of the Qur'an using Machine Learning

Abstract

The Qur'an is the Muslim holy book, consisting of 30 juz and 114 surahs of varying length and number of verses. Reading the Qur'an involves special techniques to understand and read with similarity or consistency to the verse being read. In the digital era, technology enables the development of applications that support the learning and analysis of Qur'anic recitation. This research aims to design and implement a REST API to predict Qur'an recitation using a machine learning (ML) model. This API accepts voice recordings from users and provides output in the form of an assessment of the similarity of their recitation to the desired verse. Using FastAPI and pre-trained models such as Wav2Vec2, the system can translate audio into text with fairly good accuracy. Experimental results show a word error rate (WER) of 30%, which indicates the need for further improvement but is sufficient in the experimental context. The technology is useful as a self-learning tool for the Qur'an, but it does not replace the role of the teacher. Future research should focus on improving model accuracy and integrating more user-friendly features.